On Tue, Jul 17, 2018 at 1:43 PM, Miro Hrončok <mhroncok(a)redhat.com> wrote:
On 17.7.2018 14:16, Nico Kadel-Garcia wrote:
> On Mon, Jul 16, 2018 at 6:18 PM, Charalampos Stratakis
> <cstratak(a)redhat.com> wrote:
>> ----- Original Message -----
>>> From: "R P Herrold" <herrold(a)owlriver.com>
>>> To: "Development discussions related to Fedora"
>>> Sent: Monday, July 16, 2018 8:57:11 PM
>>> Subject: Re: Intent to orphan Python 2
>>> On Mon, 16 Jul 2018, Miro Hrončok wrote:
>>>> On 23.3.2018 12:23, Petr Viktorin wrote:
>>>>> tl;dr: Unless someone steps up to maintain Python 2 after 2020, we
>>>>> to start dropping python2 packages now.
>>> tl;dr: --- that statement by itself overlooks the obvious.
>>> Not ALL packages become unsupported that first day of that
>>>>> Python 2.7 will reach end of upstream support on 1st of January,
>>>>> after almost 10 years (!) of volunteer maintenance.
>>> Not to be too direct about this, but isn't the RHEL 6 primary
>>> maintenance date (through 2020 11 30) a closer maintenance
>>> depot to look at and to compare against ?
>> I don't see how that relates to Fedora. Could you elaborate on what you
> EPEL. Many of us use EPEL, with components from Fedora backported to
> our working environments. It's been an invaluable resource. Me? I just
> got a good look at openstack,, as well, which solved a *lot* of
> problems for me trying to bring some modules for communicating with
> proprietary data appliances into a RHEL environment. It's part of why
> so many Python modules have bothered to maintain Python 2 and Python 3
I hear you. I just don't understand what our action shall be according to
you. Having python2 in Fedora might indeed be beneficial to old EPELs (and
RHELs). But it shall not be an excuse to have thousands of modules packaged
and supported because some of them might (or might not be) also present in
EPEL. You can even use python3 in EPEL and call it a day.
I was explaining why reasonable people involved in Fedora would care.
I'm not saying never do it. It also happened with the perl 4 to perl 5
upgrade, for those of us who remember that, and when apache 1.x became
httpd 2.x, and when Ruby updated to 2.x. Parallel support is possible
and sometimes necessary.
What's happening now with Python has been pretty good, and I applaud
the maintainers who've been forwarding Python 2 modules and the
occasionally messy but overall effective logic of building parallel
python2 and python3 modules. Python 2 *does* need to be replaced as
the default. And I think at some point the logic will need to be
reversed, that "with_python2" becomes the flag needed for building
those older package, instead of the current "with_python3". That is
going to be a pain in the *fundament* to port.
I (and now I'm speaking strictly for myself, other Fedora Python
might have different opinion) won't spend my free time to maintain something
I don't like just to support your commercial work. Will you? I don't have
enough resources in my paid time to support Python 2 in Fedora **on the
current scale**. That's what this topic was all about. Reduce the cruft, so
we can keep it and support it in our paid time to support both commercial
and non-commercial work. If not reduced, we cannot do that.
Your logic is sound. I publish patches to Fedora authors out of
enlightened self interest for my paid work with RHEL and CentOS, and
occasionally for the challenge of getting stuff into the bleeding edge
systems. When RHEL 8 comes out, I hope it will be Python 3 based and
I'll have the tools to mostly ignore Python 2, fostered by the good
work happening in Fedora.